Continue.dev Post-Acquisition Final Release Migration Path for Existing Hub Users
Continue.dev’s acquisition by Cursor has reached its endpoint with the final 2.0.0 release of the VS Code extension, CLI, and JetBrains plugin. For developers who built workflows around Continue’s Hub infrastructure, this release marks a hard transition: the Hub is gone, authentication is removed, and the configuration model has shifted to explicit model definitions. This guide documents the final release specifics, what changed, and how to migrate your setup to maintained alternatives.
Who This Migration Path Is Wrong For
This guide assumes you are an existing Continue.dev user with saved configurations who needs to preserve functionality. It is not the right path if you:
- Require managed team authentication or centralized billing for AI coding assistants
- Depend on cloud-hosted model routing without maintaining your own API keys
- Want a fully hands-off, commercially supported IDE integration with guaranteed uptime
- Are starting fresh and have no legacy Continue.dev configurations to preserve
If you fit those profiles, evaluate commercially maintained alternatives directly rather than attempting to reconstruct Continue’s final open-source release.
What the Final 2.0.0 Release Changed
The vendor’s GitHub releases page documents the following changes for version 2.0.0, released June 19:
| Feature | Status in 2.0.0 | Impact on Existing Users |
|---|---|---|
| CLI-install banner | Removed | No longer prompts new users to install CLI |
| Generate Rule feature | Removed | Custom rule generation workflows break |
| Onboarding / new-config template | Switched to explicit model definitions | No Hub slugs; manual model setup required |
| Deprecation banner export link | Updated | Points to current migration resources |
| Anonymous telemetry | Removed | No data collection, but also no usage insights |
| Authentication | Removed | No login state; all models require direct API keys |
The vendor documentation states this release was a “polished” final version that also included bug fixes, building on the prior codebase.
What “No Hub Slugs” Means for Your Saved Configuration
Prior to 2.0.0, Continue.dev’s Hub used shorthand identifiers—”slugs”—that mapped to preconfigured model endpoints. A slug like claude-3-sonnet or gpt-4o would resolve to the appropriate provider, endpoint URL, and context-window settings without user-visible configuration.
In 2.0.0, these slugs are eliminated. The onboarding flow and config.yaml template now require explicit model definitions: you must specify the provider name, model identifier, API endpoint (if not default), and your own API key for every model you use.
For existing users with saved Hub configurations, this means:
- Any config referencing Hub slugs will fail to resolve those models
- Previously working setups will show errors or silent failures on model calls
- You must reconstruct each model entry with full provider specifications
The vendor’s documentation notes that config.yaml replaces the deprecated config.json format, and that some settings previously stored in config.json have been auto-migrated to User Settings. However, Hub slug references have no automatic translation path—you must manually rebuild.
Back Up Your Configuration Before Any Changes
Before proceeding with migration or reconfiguration, back up your existing Continue.dev setup.
- Locate your configuration directory:
- VS Code:
~/.continue/(macOS/Linux) or%USERPROFILE%\.continue\(Windows) -
JetBrains: project-level
.continue/directory or IDE-specific config path -
Copy the entire directory to a dated archive:
cp -r ~/.continue ~/.continue-backup-$(date +%Y%m%d) -
Export any workspace-specific assistant definitions if you used the YAML-based assistant feature (documented in the vendor’s configuration reference)
-
Document your current model usage: which slugs you relied on, approximate token volumes, and any custom slash commands or context providers
This backup is irreversible protection. The 2.0.0 update modifies configuration handling, and without a snapshot you cannot reconstruct your pre-migration state.
Migrating to Explicit Model Definitions (2.0.0-Native Path)
If you remain on Continue.dev 2.0.0, follow the vendor’s documented configuration format. The config.yaml structure requires explicit model blocks:
Example based on vendor documentation schema
models:
– name: my-claude-instance
provider: anthropic
model: claude-3-5-sonnet-20241022
apiKey: ${ANTHROPIC_API_KEY}
- name: local-ollama
provider: ollama
model: codellama:7b
apiBase: http://localhost:11434
Key differences from Hub-based config:
| Hub Era (pre-2.0.0) | 2.0.0 Explicit Model |
|---|---|
model: claude-3-sonnet (slug) |
provider: anthropic + model: claude-3-5-sonnet-20241022 |
| Automatic API endpoint resolution | apiBase required for non-default endpoints |
| Implicit authentication via Hub session | apiKey or environment variable required per model |
| Centralized model availability | Per-model enable/disable in your config only |
The vendor’s configuration reference documents additional fields for context window, system message templating, and capabilities flags. Check the current documentation for your specific provider’s requirements.
Export Path: Migrating to Alternative Tools
Continue.dev 2.0.0 is a terminal release with no ongoing development. The updated deprecation banner points to migration resources. Below are documented translation paths for three alternatives.
Option 1: Cursor (Acquirer)
Cursor is the commercial entity that acquired Continue.dev. Third-party reviews note that Cursor provides substantial quality advantage for complex refactoring through its Claude model access, with a more polished codebase indexing implementation than Continue.dev’s open-source equivalent.
Configuration translation:
- Cursor uses its own
cursor/settings directory and.cursorrulesfile for project-specific behavior - No direct import of Continue.dev
config.yamlorconfig.json; manual reconstruction required - API key management is built into Cursor’s settings UI rather than file-based configuration
Doc-vs-reality drift to flag: The vendor’s acquisition messaging states the “open-source codebase remains freely available,” but Cursor itself is a separate commercial product with its own pricing structure. The free availability refers to the historical Continue.dev code, not Cursor’s current offering.
Option 2: TabbyML (Self-Hosted Open Source)
TabbyML is documented as a self-hosted AI coding assistant alternative. For teams seeking to preserve Continue.dev’s open-source, local-LLM-capable model, TabbyML represents the closest architectural match.
Configuration translation:
- TabbyML uses a
config.tomlfor server configuration and VS Code extension settings for IDE integration - Model definitions are server-side rather than client-side; you configure the TabbyML server with model endpoints, then point the IDE extension at your server
- No automatic migration from Continue.dev’s
config.yaml; manual model redefinition required
Doc-vs-reality drift to flag: Third-party benchmarks include TabbyML in comparison sets, but the vendor’s own documentation does not provide explicit migration steps from Continue.dev. Expect to rebuild configuration from scratch.
Option 3: GitHub Copilot
GitHub Copilot is the most commercially established alternative. Third-party reviews state that Copilot’s inline completion is best here since it understands the GitHub ecosystem deeply.
Configuration translation:
- Copilot uses GitHub account authentication rather than file-based API key configuration
- No equivalent to Continue.dev’s
config.yaml; settings are managed through VS Code’s built-in Copilot panel orsettings.jsonentries - Custom model selection is limited to GitHub’s offered model roster; no arbitrary provider endpoint configuration
Doc-vs-reality drift to flag: One third-party review notes that Continue.dev is the only tool among mainstream AI code assistants that simultaneously satisfies “Open Source, Local LLM, and JetBrains Support.” GitHub Copilot satisfies none of these three criteria. If any of those properties were essential to your Continue.dev usage, Copilot is not a functional substitute.
Configuration Export Checklist
Use this checklist to ensure complete migration coverage:
- [ ] Back up
~/.continue/or equivalent directory (dated archive) - [ ] Inventory all Hub slugs used in previous configuration
- [ ] Obtain direct API keys for each model provider previously accessed through Hub
- [ ] Rebuild model definitions in target tool’s configuration format
- [ ] Recreate custom slash commands or context providers (no automatic translation)
- [ ] Verify IDE extension version compatibility with target tool
- [ ] Test code completion and chat functionality on a non-critical project first
Security and Applicability Boundaries
This guide documents configuration migration based on publicly available vendor documentation. It is not security advice. Consider the following boundaries:
- Removing anonymous telemetry and authentication in Continue.dev 2.0.0 improves privacy but eliminates security telemetry that might detect misuse
- Storing API keys in
config.yamlor environment variables shifts security responsibility to your local machine and secret management practices - Self-hosted alternatives like TabbyML introduce infrastructure security considerations not present in managed services
Evaluate your threat model independently before selecting a migration target.
Final Release Status and Ongoing Support
The vendor’s GitHub repository documents 2.0.0 as the “final release.” No subsequent releases are planned. The repository remains available as “a foundation for others,” per the vendor’s acquisition statement, but this is source code availability, not product support.
For developers who built production workflows on Continue.dev’s Hub infrastructure, the executable path is: back up, inventory dependencies on Hub-specific features, select a maintained alternative based on which Continue.dev properties you actually used, and rebuild configuration in that tool’s native format. The 2.0.0 release provides a stable endpoint for those who wish to continue using the frozen codebase, but the removal of authentication, telemetry, and Hub routing makes it a self-supported option requiring direct API management for every model interaction.
For readers prioritizing privacy-preserving AI assistance with self-hosted or enterprise deployment options, Tabnine offers a privacy-first track that may align with the local-LLM workflow previously supported by Continue.dev’s open-source architecture.
Version-Anchored Reference Summary
| Tool | Version Referenced | Documentation Source |
|---|---|---|
| Continue.dev | 2.0.0 (final) | Vendor GitHub releases page |
| Continue.dev config format | config.yaml (current), config.json (deprecated) |
Vendor configuration reference |
| Cursor | Current commercial offering | Third-party reviews |
| TabbyML | Included in benchmark comparisons | Third-party benchmark collections |
| GitHub Copilot | Current commercial offering | Third-party reviews |
